Exploratory data analysis of Asia-Pacific political incident and conflict trends

1. Introduction and broad trends

This report documents observed conflict trends in the Asia-Pacific region over the past ten years, exploring trends relating to events, fatalities, conflict types and conflict actors using ACLED data¹.

This report is in line with ACLED’s metholodology² and its categorisation of conflict events, which encompasses protests, riots, battles, explosions and remote violence and strategic developments (such as arrests of key figures).

Figure 1 below displays cumulative conflict events over the past 10 years (2014-2023).

Major conflict-affected countries in the region are all quite clear: Afghanistan, Pakistan, Bangladesh, India, Thailand, Philippines and Myanmar, as are the “hotspots” within them (e.g. Kashmir) that account for some of these countries being on the list whilst not necessarily being considered to be in “in armed conflict” at the country level.

Figure 1

However, both the magnitude (number of events or fatalities, as displayed above) and the severity (events and fatalities per capita) provide important perspectives on conflict: India may have a very large number of conflict events, but as can be seen from Figure 2 below, on a per-capita basis (in this case the denominator per 100,000 persons), it sits sin the middle of the pack.

The plot below is of the number of events per 100,000 persons per annum against the number of conflict fatalities per 100,000 persons per annum. The number of events is represented by the size of each point and the colour represents the total conflict fatalities experienced by each country over the past 10 years.

Figure 2

Make this a gif?

Whilst the immensity and diversity of the Asia-Pacific demand exceptions to every rule, we can identify a central tendency (the solid blue line) that most countries in the region seem to sit on. On one end of the line are disproportionately quiet autocracies (or their approximations) and on the other are countries where violence has spun out of state control.

Myanmar and Afghanistan have separated themselves from this central curve as a consequence of open warfare. Philippines, Papua New Guinea and Pakistan are all at risk or were at risk of doing the same in the past 10 years. The blue dotted reference lines were chosen based on a visual inspection of the data.

At the bottom of the plot, on either side of the reference line at x = 1, we see on one side, autocracies and on the other, democracies. These democratic countries have exited the central tendency as well. In these countries, an increase in political participation and conflict is not associated with an increase in fatalities.

Furthermore, all have populations who actively participate in the political process – even the least protest-inclined of this group, Japan, has higher rates of participation (i.e. number of events per 100k) than Indonesia or Malaysia, or even pre-2024 Bangladesh.

ACLED does not capture all conflict. The number of fatalities per 100,000 here is for comparison amongst countries within the ACLED dataset. Were we looking for the actual mortality rate, there are preferred sources, which put the mortality rate in Afghanistan from “collective violence and legal intervention” at 122 per 100,000 in 2021. The value for ACLED in Afghanistan in 2021 was 106 per 100,000. The homicide rate in Southeast Asia in four per 100,000, the road traffic death rate is 16.

Over the course of this document, as we explore trends and drivers of conflict in the region, it is important to recall that there is no optimal amount of political conflict. Too little indicates suppression and too much, widespread dissatisfaction and, in some cases, open revolt amongst the populace.

But until we identify why and how countries can exit the central tendency and join the peaceful countries in the bottom-right corner, the majority of countries in the region will oscillate between upheavals and crackdowns, between disarray and abuses of state power.

To paraphrase Martin Luther King, Jr., without justice, there is no peace, only order.

1.1 Conflict events and fatalities (summary tables)

Looking back at a 10-year horizon (note that ACLED used to monitor a smaller set of countries) in Table 1, although the number of events is much higher than 10 years ago, the majority of countries have seen declines in the number of fatalities associated with political conflict. However, most of these countries have also seen increases in the number of events, indicating increased that whilst there is more conflict, it is less bloody than before (except in Myanmar).

Table 1: 10 years of conflict: comparison 2014 and 2023

Country	Events 2014	Events 2023	Events Diff%	Fatalities 2014	Fatalities 2023	Fatalities Diff%
Myanmar	899	13,232	1,371.86	501	16,060	3,105.59
Nepal	250	2,490	896.00	6	10	66.67
Sri Lanka	158	894	465.82	7	7	0.00
Bangladesh	1,123	2,693	139.80	391	337	-13.81
Pakistan	4,201	8,536	103.19	5,419	2,333	-56.95
Thailand	1,304	608	-53.37	424	57	-86.56
Vietnam	47	17	-63.83	29	10	-65.52
Cambodia	270	90	-66.67	23	0	-100.00
Total	8,253	28,562	246.08	6,800	18,814	176.68

Across a five-year horizon, with more countries included in the comparison, the number of events and fatalities is down in most countries, though regional totals are boosted by Myanmar.

Afghanistan, Vietnam and Hong Kong saw the largest declines in political conflict. Myanmar, Indonesia, Iran and Bangladesh have each experienced a significant increase in unrest in the past five years. As mentioned before, though countries like South Korea and Japan also large increases in political conflict, these activities were not associated with any deaths.

Table 2: Five years of conflict: comparison 2019 and 2023

Country	Events 2019	Events 2023	Events Diff%	Fatalities 2019	Fatalities 2023	Fatalities Diff%
Myanmar	1,815	13,232	629.04	1,512	16,060	962.17
Indonesia	1,644	3,570	117.15	210	214	1.90
North Korea	19	37	94.74	9	12	33.33
Iran	2,602	4,066	56.26	478	214	-55.23
Bangladesh	1,738	2,693	54.95	391	337	-13.81
Japan	1,299	1,792	37.95	0	0	0.00
South Korea	5,493	6,321	15.07	0	0	0.00
Pakistan	7,686	8,536	11.06	1,157	2,333	101.64
Thailand	568	608	7.04	174	57	-67.24
India	23,672	22,865	-3.41	1,529	1,213	-20.67
Sri Lanka	954	894	-6.29	304	7	-97.70
Mongolia	94	87	-7.45	0	0	0.00
Cambodia	98	90	-8.16	5	0	-100.00
Nepal	3,242	2,490	-23.20	49	10	-79.59
Malaysia	224	155	-30.80	15	0	-100.00
Philippines	1,901	1,251	-34.19	1,706	855	-49.88
China	2,516	1,647	-34.54	13	9	-30.77
Vietnam	58	17	-70.69	1	10	900.00
Afghanistan	8,373	1,465	-82.50	23,413	1,125	-95.19
Hong Kong	1,235	96	-92.23	1	0	-100.00
Total	65,235	71,914	10.24	30,969	22,456	-27.49

The increase in the number of conflict events – especially if they were primarily due to protests – is not an immediate cause for concern.

To provide a bit more perspective on these country totals, and given that declines or increases happen for a multitude of reao

Increases in the number of conflict events – especially if they were primarily due to protests – is not an immediate cause for concern. South Koreans have managed to enact several changes through their protests (like the removal of President Park Geun-hye), as have Malaysians and Nepalese. Similarly, a decrease in conflict events only indicates stabilisation or suppression, not that the underlying issues have been resolved.

To provide an additional layer of perspective on Table 1 and Table 2, both the number of events and the number of fatalities are plotted below in Figure 3.

For most countries, conflict fatalities are down or flat. This is in line with what we have observed in the ten-year and five-year horizons.

Afghanistan, China, Vietnam, Cambodia, Laos, Philippines and the Solomon Islands have each seen a drastic fall in conflict events and fatalities over the past 10 years. Fatalities in Pakistan and Bangladesh were trending upwards, at the end of 2023 (2024 has not yet concluded and for this comparison, we have elected to use only full years of data).

However, looking at the dark blue lines, indicating the number of events, it becomes clearer that the remaining countries all experience periodic spikes and troughs in conflict, periods of destabilisation and tightening. Clear examples of this are Iran and Thailand, where the waves of protest movements or elections are immediately visible. As are the effects of the state re-exerting control.

If we were to extrapolate from the example of Iran: the population there seems to be severely dissatisfied with their leaders and are expressing it. There were two clear waves of protests and while numbers might currently be on the downtrend from last year, the next protest movement will likely be even more destabilising than the last.

Figure 3

A number of territories – China, Nepal, Hong Kong, Japan, Mongolia and South Korea – have very low or 0 fatalities from political incidents. We note that the line for fatalities in Figure 3 above in these countries is flat. We venture that whilst countries like China are “orderly”, they not necessarily peaceful: its trendlines are most similar to those of Vietnam and Cambodia (where fatalities are low because events are suppressed) rather than Japan or South Korea (where the fatalities line stays flat no matter how many events there are). It is hypothesised that its conflict fatalities being flat might more be due to prolonged suppression.

1.2 Event type streamgraphs

To add further context to our exploration of regional conflict trends, let us now consider event types. ACLED³ categorises conflict events into protests, riots, battles, explosions and remote violence and strategic developments (such as arrests or new legislation). These are further still into 25 sub-types.

The charts below are streamgraphs, where values over time – in Figure 4 below, the number of quarterly conflict events – are represented by their respective area. Though there are negative values shown, the Y-axis values are absolute i.e. if a certain point in time is +400 and -400, there were a total of 800 events. Streamgraphs are employed here despite this because spikes in violence (from elections, for instance) are much more visible than a traditional area graph.

Figure 4 shows the number of political incidents per quarter broken down by event type, sorted in order by the number of political incidents.

It should be noted that in “stable” countries, the flavour of political conflict consists mainly of protests (which indicate widespread dissatisfaction but also that not all nonviolent means have been exhausted).

The Thai people manifested their dissatisfaction has manifested in 2014 and in 2021. However, the Thai state responded by firmly clamping down on its citizenry in both instances, as can be seen from the swift decrease in events. This approach does not address the core issues behind the protests. Notably, the share of protests compared to violent political incidents seems to be worsening, possibly indicating that more actors (the state included) believe that nonviolent solutions have been exhausted.

Bangladesh seems to have an outsized number of riots. Whilst protests are a natural feature of any democracy, riots are less so, given their disruptive and violent nature. Bangladesh seems to have a higher ratio of riots to protests than any other country except Papua New Guinea. This speaks to the inability of the state to maintain order and the ease of inciting the population to violence (which speaks to the existence of many instances of injustice and hurt to inflame). Political incidents in Bangladesh (as we will explore in later section) had been falling until they increased around 2018, around the time of the first quota reform protests.

Figure 4

Myanmar and Afghanistan bear all the patterns of war. But what is surprising is the extent to which Philippine civil society has deteriorated due to damage from the drug war and the resulting political violence. When compared to other countries, their share of protests is very low, lower even than Cambodia. Even as the drug war operations ceased and killings began to drop off sharply, the flavour of political unrest in the Philippines seems deeply changed, with its participatory democracy suffering massive setbacks. They are, after all, the country of both the EDSA revolution and the dictator Marcos.

Let’s take a closer look at some countries of interest and narrow down from a ten-year horizon to a five-year one:

Figure 5

From the dramatic decrease in political activity starting in 2020, it would appear that the Vietnamese state has seized the opportunity to presented by COVID to advance state authority and solidify its hold on the country. China saw a more modest reduction in political incidents, and is undergoing an increase in unrest related to labour and wage issues.

Indonesia is has also seen a large increase in political incidents driven by economic woes, increasing state authoritarianism and violations of human rights and environmental and land rights.

In post-genocide Myanmar, with the longstanding civil war with ethnic and communal militias in the background, the population exploded in protests against the Tatmadaw’s coup. However, nonviolent means were soon exhausted and the country is now in full civil war.

With the victory of the Taliban, Afghanistan has seen a 96% reduction on conflict fatalities and a 83% drop in conflict events since 2019 levels. However, civil society has by no means recovered, nor does Afghanistan bear any of the hallmarks of a participatory democracy.

2. Protests

The most common form of political conflict is to protest. Protests are a non-violent form of political conflict which manifest when all legal and administrative avenues have been exhausted. Widespread protests indicate a “hole” in governance, given that such protests would only be possible if a significant proportion of the population not only agreed with them but also felt that the issue had gone unaddressed long enough or is egregious to manifest their outrage in public.

2.1 Common protest topics

Figure 6 below shows the (log of) odds that a word will appear in post-COVID event descriptions for protests events in that country. For example, the word with the highest odds of appearing in protest events in India is BJP.

This document does not pretend to speak for the people, however, the charts below do provide a very useful overview of what people in each country are most concerned with. To continue with India, the top end of its chart is dominated by minority rights and rural economic issues, with the incumbent party firmly at the top, given that these protests stemmed from their policies. Instead of focusing of the individual odds of each word, it is better to take the list of words as a whole, in aggregate.

Place names have been largely filtered out, countries below left with many place names in their protest descriptions tend to have experienced widespread protests over singular topics. An example of this is Thailand, where protests seem to revolve around the singular issue of state power. Almost every other word is a place name where protests concerning state power took place. Myanmar’s protest descriptors are similar as well, given that they too were preoccupied with their own coup. This is also the same for South Korea, though that may also be due to over-centralisation in Seoul.

Figure 6

The Philippines has less protests per capita than Afghanistan. Protesters there are not in an enviable position if they believe protesting is riskier than it is in Afghanistan.

South Korea, Hong Kong, the Maldives and Nepal all experienced significant protest events. New Zealand is currently experiencing a political crisis, but people there might just be more politically active in general, given that in addition to Maori rights, they seem to be fairly responsive to geopolitical events.

The people of Iran, in spite of their government being party to numerous contentious geopolitical conflicts, seem to be most concerned about economic issues as well as basic human rights. And in Sri Lanka, tensions from the civil war continue to manifest in society.

2.2 Use of force against protesters

This section is concerned with protests where individuals who are peacefully protesting are targeted with violence, including that which results in serious injuries or death. Figure 7 charts the countries where protests are most likely to be met with violence.

More than one-fifth of all protests in the Maldives and Vietnam were met with violence and/or excessive force. Criticisms of the Maldives for restrictions on freedom of expression and peaceful assembly are validated by the data. The next four countries below it (Vietnam, Cambodia, Afghanistan, Myanmar) either have massive state apparatuses dedicated to suppression of free expression or were or ar at war.

violence ratio pre- and post-covid

Figure 7

The level of suppression is indicative of both the priorities of the state (how committed it is to free expression) as well as the ability of the state to enforce its authority. The New Zealand and Australian government’s responses to protests was swift and forceful, especially during COVID. On the other hand, South Korea and post-COVID Japan both seem to be much more permissive towards protesters. However, this does not mean that the government of Pakistan is more permissive to protesters than that of Australia, just that the authority of the state there is too weak to do.

Likewise, the number of protests met with violence in Bangladesh decreased from about 10% pre-COVID, to just over 5% post-COVID whilst the number of total protests has doubled: given that the quota reform movement and the resulting government response there began before COVID, the Bangladeshi state was likely unable to intervene in any more protests than it already was doing.

3. Country spotlights

3.1 Myanmar

The current conflict is very linear and concerned with the control of the road network. The highway network was constructed to link military garrisons, important natural resources (gems, ores, oil and gas), border crossings and, now, narcotics factories important to funding the war effort. The Tatmadaw have never been in full of Myanmar’s territory and their ability to project power is now deeply compromised and, consequently, so is their economy.

This has given rise to the very linear pattern of violence as both sides contest for control over the road network. For the rebels, they have stated that “all roads lead to Naypyitaw” (the gold diamond in the middle of the country) are have been trying to enclose the military and limit its movements.

Overland trade with China has fallen by 41% and by 87% with Thailand; the border crossing with India remains closed. Rebel groups are now preparing to siege Sittwe and Mandalay, the last major garrison before Naypyitaw.

Figure 8 plots the Myanmar civil war over its road network (dark blue). Conflict events are in red, with the sizes corresponding to the number of fatalities. Gold squares are Tatmadaw regional commands and blue squares are border crossings.

Figure 8

In the northeast, Lashio has fallen, drastically lowering the amount of conflict in the area. And Muse has lost five out of its ten border gates with China. In the southeast Loikaw has fallen (not a major command post, but a city at major arterial intersection in the road network) and so has the crossing with Chiang Mai at Myawaddy.

In the northwest, India has closed the border crossing at Tamu (in line with the Tatmadaw’s wishes, since they no longer control it); and we also see intensification of conflict in Rakhine in preparation for the battle of Sittwe. With the northern garrisons isolated and cut off, rebels hold much of Monywa town itself, isolating the garrison there as the main front encircles and prepares to invade Mandalay.

The military no longer has the ability to push beyond the positions they already hold. If the rebels can defeat the Mandalay garrison, the Tatmadaw will have lost. However, the remaining garrisons are entrenched and manned by hardened troops. With reference to the inset line graph of fatalities over time in Figure 8, although overall fatalities seem to be decreasing, battles seem to be getting bloodier as the conflict gets more concentrated as spikes in fatalities in 2024 matched the highs set in 2022.

However, should the fall of the Tatmadaw seem imminent, all attempts must be made, through all channels, to prevent a free-for-all. With reference to MIMU ethnolinguistic map of Myanmar (Figure 9) below, no [modern] government has exerted territorial control over the entirety of Myanmar.

Figure 9

Areas predominantly settled by ethnic minorities are strongholds of anti-Tatmadaw forces and have been for decades. Forces in these areas are converging on the yellow area in the middle of the country, populated primarily by the Bamar majority.

Given the disparate armed groups present, and the massive proliferation of conflict actors in post-coup Myanmar, and that previous attempts at peace and reconciliation in Myanmar, even under the Aung San Suu Kyi administration, were half-hearted and not well-received.

Figure 10

The rebels and the National Unity Government have sought to isolate military garrisons from each other.

The plots above are of the Myanmar civil war, with events (red) plotted over the road network (blue). Gold squares are Tatamadaw regional commands and the blue squares are border crossings.

Within Myanmar itself, with reference to the maps below, the fighting seems to have come to a very critical juncture. At first glance, the fighting seems to have lessened in 2024, but with reference to the line graph of monthly fatalities, we know that the fighting has instead been concentrated.

The rebels and the National Unity Government have sought to isolate military garrisons from each other; and the Tatmadaw needs to maintain its supply lines to its major command centres (as they protect border crossings, various mineral mines and other resources, such as methamphetamine factories) as well as try to project power into rebel-held territory.

3.1.1 “Deadliest” actors

The plot below details the most lethal actors by country: these are the actors involved in events associated with the highest number of fatalities. The x-axis is the estimate for the number of fatalities from each event each actors is involved in.

Additionally, actors related to or originating from Myanmar are highlighted in red in the plot below. The widespread nature of protests, riots and violent incursions related to Myanmar. Even though armed conflict in Myanmar is largely contained within its borders, a war of this size is destabilising for the region as a whole.

The ILO noted that there were 2.2 million Myanmar migrants in Thailand and Malaysia alone in 2019. It is unclear how much more that number has grown since the start of the civil war.

The Philippine police is more lethal than anti-drug vigilantes and the New People’s Army. Similarly, in China, civilians seem to be involved in more fatalities per event than even the police, or rioters, indicating problems with violence against civilians (as fatalities from an incident where one party were civilians and the other combatants).

South Korea here should be counterpoint to the Philippines, Vietnam and China, especially in terms of the lethality of state actors. In spite of how much Koreans protest, neither their protesters nor their state actors are amongst those who have suffered the most casualties.

Appendices

Fatalities streamgraph

From Figure 11, sorted in order of number of fatalities from political incidents per quarter, it is clear that Afghanistan, Myanmar, Pakistan and Philippines have experienced severe political unrest, ranging from war, terrorism and political killings. Conflict in Afghanistan and Philippines has fallen significantly since its peak, but Myanmar is still in full conflagration and Pakistan is on the upswing:

Figure 11

Protest network graph

It appears that Indonesia, Iran, Nepal and Cambodia each have experienced much cyclical political violence related to elections and mass protest movements. Though in Cambodia, the deaths associated with protests in 2018 might have served as a strong deterrent.

In this section, we explore the concerns of protesters in the Asia-Pacific. Below is a network graph of event descriptions of protests in the Asia-Pacific in the past 10 years. The links (edges) between each of the words indicate the strength of the relationship (transparency) and the number of times this word pair has occurred (thickness). Only the most common word pairs with correlations above 0.2 are included (for legibility).

The main topics identified from the graph include (cross-border issues are in bold):

• Contentious territories: Myanmar, Palestine, Korea, Manipur, Kashmir, Fukushima

• Legislation and state overreach: Citizenship amendment, Lese majeste, Oli (Nepal), Arrest, Money laundering

• Economic concerns and service delivery: Unions, Prices, Arrears, Pensions, Hikes, Load shedding, Farms, Outages

• Coronavirus

These can be thought of as the prime destabilising factors for the region: in some cases, they have devolved into open conflict. Peace actors should seek the just resolution of issues and ensure that nonviolent options always remain open to people.

Regional tensions, state overreach and bread and butter issues are the issues that people in the Asia-Pacific protested the most about. Though inflation seems to have been arrested, the poverty rate in these countries should be monitored very carefully.

Figure 12

Footnotes

1. The primary dataset used is the Armed Conflict Location and Event Data (ACLED). The primary reason for privileging this source over UCDP’s Georeferenced Event Dataset is because ACLED has a wider diversity of sources and records a wider range of conflict events.

2. The ACLED codebook contains all the documentation, including coding decisions and explanations of all recorded variables

3. The ACLED codebook contains all the documentation, including coding decisions and explanations of all recorded variables